NRT capacity: repeated mem_overcommit_exceeded blocks deploys and machine starts

For two consecutive days (July 12–13, 2026), we have been unable to reliably deploy or start a Fly Machine in nrt.

governor policy blocked start: hard gate failed: mem_overcommit_exceeded (memory allocation would exceed max overcommit)

The public status page currently reports NRT as operational, but the same capacity failure has affected us on two consecutive days.

Questions:

  1. Is NRT currently experiencing host capacity constraints that are not reflected on the public status page?
  2. Is there a way to move a volume-backed Machine to another NRT host without migrating the application to another region?

Thank you!

the nrt region as a whole seems to have a decent amount of spare capacity, so this is likely one single host at capacity. you can fork the volume and create a new machine with the new volume, which will land on a different host.